JAL-1705 refactoring etc for fetching Ensembl --> Uniprot

[jalview.git] / src / jalview / ext / ensembl / EnsemblCdna.java

diff --git a/src/jalview/ext/ensembl/EnsemblCdna.java b/src/jalview/ext/ensembl/EnsemblCdna.java
index 757b3c8..856be74 100644 (file)
--- a/src/jalview/ext/ensembl/EnsemblCdna.java
+++ b/src/jalview/ext/ensembl/EnsemblCdna.java
@@ -1,17 +1,61 @@
 package jalview.ext.ensembl;
 
-import jalview.ext.ensembl.SeqFetcher.EnsemblSeqType;
+import jalview.datamodel.SequenceFeature;
+import jalview.io.gff.SequenceOntologyFactory;
+import jalview.io.gff.SequenceOntologyI;
+
+import java.util.Arrays;
+import java.util.List;
 
 import com.stevesoft.pat.Regex;
 
+/**
+ * A client to fetch CDNA sequence from Ensembl (i.e. that part of the genomic
+ * sequence that is transcribed to RNA, but not necessarily translated to
+ * protein)
+ * 
+ * @author gmcarstairs
+ *
+ */
 public class EnsemblCdna extends EnsemblSeqProxy
 {
+  private static final List<String> CROSS_REFERENCES = Arrays
+          .asList(new String[] { "Uniprot/SWISSPROT", "Uniprot/SPTREMBL" });
+
+  /*
+   * accepts ENST or ENSTG with 11 digits
+   * or ENSMUST or similar for other species
+   * or CCDSnnnnn.nn with at least 3 digits
+   */
+  private static final Regex ACCESSION_REGEX = new Regex(
+          "(ENS([A-Z]{3}|)[TG][0-9]{11}$)" + "|" + "(CCDS[0-9.]{3,}$)");
+  
+  /*
+   * fetch exon features on genomic sequence (to identify the cdna regions)
+   * and cds and variation features (to retain)
+   */
+  private static final EnsemblFeatureType[] FEATURES_TO_FETCH = {
+      EnsemblFeatureType.exon, EnsemblFeatureType.cds,
+      EnsemblFeatureType.variation };
 
-  public EnsemblCdna() throws Exception
+  /**
+   * Default constructor (to use rest.ensembl.org)
+   */
+  public EnsemblCdna()
   {
     super();
   }
 
+  /**
+   * Constructor given the target domain to fetch data from
+   * 
+   * @param d
+   */
+  public EnsemblCdna(String d)
+  {
+    super(d);
+  }
+
   @Override
   public String getDbName()
   {
@@ -27,13 +71,55 @@ public class EnsemblCdna extends EnsemblSeqProxy
   @Override
   public Regex getAccessionValidator()
   {
-    return new Regex("((ENST|ENSG|CCDS)[0-9.]{3,})");
+    return ACCESSION_REGEX;
+  }
+
+  @Override
+  protected EnsemblFeatureType[] getFeaturesToFetch()
+  {
+    return FEATURES_TO_FETCH;
+  }
+
+  /**
+   * Answers true unless the feature type is 'transcript' (or a sub-type in the
+   * Sequence Ontology).
+   */
+  @Override
+  protected boolean retainFeature(SequenceFeature sf, String accessionId)
+  {
+    if (isTranscript(sf.getType()))
+    {
+      return false;
+    }
+    return featureMayBelong(sf, accessionId);
+  }
+
+  /**
+   * Answers true if the sequence feature type is 'exon' (or a subtype of exon
+   * in the Sequence Ontology), and the Parent of the feature is the transcript
+   * we are retrieving
+   */
+  @Override
+  protected boolean identifiesSequence(SequenceFeature sf, String accId)
+  {
+    if (SequenceOntologyFactory.getInstance().isA(sf.getType(),
+            SequenceOntologyI.EXON))
+    {
+      String parentFeature = (String) sf.getValue(PARENT);
+      if (("transcript:" + accId).equals(parentFeature))
+      {
+        return true;
+      }
+    }
+    return false;
   }
 
   @Override
-  public String getTestQuery()
+  protected List<String> getCrossReferenceDatabases()
   {
-    return "ENST00000288602";
+    return CROSS_REFERENCES;
+    // 30/01/16 also found Vega_transcript, OTTT, ENS_LRG_transcript, UCSC,
+    // HGNC_trans_name, RefSeq_mRNA, RefSeq_mRNA_predicted
   }
 
 }